The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface
Ingestion points: scripts/auto_mirror.py retrieves tweet text from the vendor-managed Twitter API proxy (ai.6551.io).
Boundary markers: Absent. Tweet content is directly interpolated into a pre-defined text template before being sent to Binance Square.
Capability inventory: scripts/auto_mirror.py and scripts/mcp_server.py possess the capability to perform network POST/PUT requests to Binance Square and write state information to a local JSON file.
Sanitization: The script performs basic sanitization on tweet text, including removing t.co URLs, stripping retweet prefixes, and collapsing multiple newlines via regular expressions.
[COMMAND_EXECUTION]: Local Command Execution for Media Processing
The skill utilizes subprocess.run to call ffmpeg and ffprobe for video duration detection and cover frame extraction. Arguments are passed as a list, which effectively prevents shell injection attacks. These operations are confined to processing media files intended for upload.
[EXTERNAL_DOWNLOADS]: Remote Asset Retrieval
The skill fetches tweet metadata from the author's infrastructure (ai.6551.io) and downloads media assets from Twitter's content delivery network (pbs.twimg.com) or user-supplied URLs. These downloads are required for the skill's primary function of mirroring social media content.
[SAFE]: Verified Service Interaction
External communications are limited to well-known services (Binance, Twitter) and the vendor's own API domain. Credentials (API keys and tokens) are managed through environment variables rather than hardcoding, following standard security best practices.

twitter-to-binance-square